Testing the predictive performance of distribution models

By V. Bahn, Brian J. McGill. 2013

245 citations recorded by the source corpus.

Subfields: Ecological Modeling, Ecology, Nature and Landscape Conservation

View Paper: https://www.semanticscholar.org/paper/4059bd3c3cd835788592c50ddb242ddacdf4acd1

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecological Modeling

In Ecology

In Nature and Landscape Conservation

Citations in the corpus

Performance evaluation and hyperparameter tuning of statistical and machine-learning models using spatial data (2018, background): “Various other ecological modeling studies conﬁrm the ﬁnding that RF is among the best performing models (Bahn & McGill, 2012; Jarnevich et al., 2017; Smoli´nski & Radtke, 2016; Vorpahl et al., 2012).”
Dynamic population mapping using mobile phone data (2014, background): “training and test sets are sampled from geographically distinct regions (6).”
Importance of spatial predictor variable selection in machine learning applications - Moving from data reproduction to spatial prediction (2019, methodology): “…shown that the commonly used random cross-validation provides considerably overopti-mistic error estimates due to the problem of autocorrelation (Bahn & McGill, 2013; Micheletti et al., 2014; Juel et al., 2015; Gasch et al., 2015; Gudmunds-son & Seneviratne, 2015; Roberts et al., 2017; Meyer et…”
Selecting predictors to maximize the transferability of species distribution models: lessons from cross‐continental plant invasions (2017, methodology): “106 Fully testing the ability of SDMs to predict species’ distribution through space or time requires an 107 independent test dataset (Bahn & McGill, 2013).”
Hypervolume concepts in niche‐ and trait‐based ecology (2018, background): “Importantly, the cross-validation can occur either on different folds of data resampled from the calibration or on wholly novel test data (Bahn and McGill 2013).”
A quantitative review of abundance-based species distribution models (2021, background): “Current abundance model evaluations examine only a limited set of statistical frameworks and the best options may be overlooked (Pearce and Ferrier 2001, Potts and Elith 2006, Oppel et al. 2012, Bahn and McGill 2013).”
Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure (2017, background): “…on random data splits (all of which will be largely consistent in underlying structure) as well as the various parametric modelling options (Table 1) will tend to underestimate predictive error (Araújo et al. 2005, Veloz 2009, Bahn and McGill 2013), leading to false confidence in model predictions.”
blockCV: an R package for generating spatially or environmentally separated folds for k-fold cross-validation of species distribution models (2018, methodology): “The package also allows division of the study area into vertical or horizontal bins of a given height or width (rectangle blocks, Figure 1b,c), as used by Wenger and Olden (2012) and Bahn and McGill (2012) respectively.”
The priority of prediction in ecological understanding (2017, background): “These metrics are notoriously bad at estimating how well a model predicts out-of-sample data (Picard and Cook 1984, Bahn and McGill 2013).”
Paintings predict the distribution of species, or the challenge of selecting environmental predictors and evaluation statistics (2018, background): “…thus confirm previous findings that it is advisable to give priority to evaluation approaches that use spatially independent occurrences located in a distant area from the calibration dataset (Bahn & McGill, 2013; Radosavljevic & Anderson, 2014, Roberts et al., 2017; Wenger & Olden, 2012).”